Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Keith
Yeah, tx, I thought so - however it is not impossible there will be a notebook design (most probably from SUN) based on Broadcom chipsets for Opteron. Not targeting the low power segment though, but folks who don't touch anything without ECC-RAM anyway and prefer to work on the same platform on any machine they boot up.
It's not exactly a volume segment, but these chaps don't mind to pay 5K for such a machine - or better said let their employers pay for it.
K.
Congratulations
borusa and i-hub-team for winning the contest and Albert_Blue
for the 4-quarter-lead.
K.
Combjelly
but you get the picture...
Frankly, no i don't. What do you mean?
K.
cd
Thanks. I have it in my notebook. It consumes juice like hell.
I should mention powerplay is not supported in the design I have.
Maybe it is in the Dell Design.
K.
cd
Thanks. I've seen a lot of excellent server-chipsets from broadcom, but none for mobiles yet. I don't see any design-synergies as well.
As for the discrete gfx, it would not make much sense for me to put a Low-power CPU onto a chipset and use the savings for grafics instead of extended batterytime for me but there could be folks looking for such a solution. Do you happne to know which discrete grafics chip is on the Inspiron 600?
K.
I_banker
While I agree.. a site like THG depends on clicks...
K.
cd
How comes you assume a Broadcom-Chipset for mobiles? (I guess it would certainly need an integrated grafics chipset for low-power notebooks, and I would be surprised to see one from Broadcom.)
K.
JJ,
for the record, just in case the revision I made this morning did not get through to you
AMD Revisions
Flash 440
Total rev 1269
Eps 0,10
Stockprices
AMD 17,5
Intel 21,5
K.
GothicChess
Ah, that is why I never came across your site probably.
Tx for the integer/fp elaborations.
K.
Although the product is not released, samples are out.
Measurements of silicon, e.g. here http://www.hkepc.com/hwdb/p4660-4.htm
strongly indicate the specs of the table I posted for you are indeed accurate. Anything else would be a huge surprise anyway. It's the same transistors, just couple of more of these.
But then, hope dies last. So these temperatures could certainly be made up as the table is and M2 is not really Prescott but rather V2. (Wunderwaffe)
Apologies four touching your innocence.
Keith
D'accord. Noble artwork. The most appealing logo i have ever seen from AMD. By a wide margin.
Centrino logo looks pretty kittenish in comparison.
K.
Well - thanks lot indeed for sharing all your valuable knowledge in this condensed manner.
K.
alan
I certainly assume all of the capacity is migrated to 2MB-process. Hard to say how much yield loss it costs. Optimistically, it is just in the ballpark of a stepping, but certainly not less than that. Initially any two-digit figure percentwise, whatever. Pretty irrelevant. More important is how quick you come back to the curve with it.
The cache thing it adresses could be more sophisticated than just a defect density issue. Somewhat along the line of variations of threshold-voltages where you either cannot make use of an array because these vary more than tolerable within the array, even using redundancy, and even if all cells would basically work at a certain Vth because of variation, or array A can be used at Vth X and array B at Vth Y but you cannot use both arrays together because delta X-Y is beyond spec. So leaving the size of the arrays unchanged and doubling it would improve things vastly. Anything like this is the most probable explanation I can find for what problems might plague Prescott more than thermal issues, but it would not allow to yield as many full-cache parts to allow the pace of migration of products you sell you see. It is nothing more than a hypothesis - but it is based on cache problems Jerry Moench lined out for the very same geometry a while ago. While Intel uses different SRAM cells and design - physics is physics. Although the attitude in Santa Clara seems to be: Physics? Who is that then?
We are Intel. And somewhat rightly so - as they are about to prove they can deal with the problem in a way almost nobody would ever notice there was a problem at all in this respect.
As for your prediction - reality check in five and a half months will reveal if its on the spot. I hope you and I will have the opportunity to discuss it further then.
K.
GothicChess
Welcome to the board, to begin with. Being a pretty lousy player myself, I love the game and sacrificed many nights and bottles to it. I never came across your site, probably because what Fritz and chessbase offer is already more than I can digest. As for the bit discussion, comp.arch might be the appropriate place to discuss it. The topic is not unknown there
For the Gallium Arsenide question, sci.engr.semiconductors might be appropriate, it is a small group but if you get an answer there it is usually well beyond what google has to offer. But I think the answers offered here might be pretty accurate as well.
While at Chess-programs, just curious - if you allow a silly question: Is FPU of any use for such kind of programs, or do these rely mainly on the integer part of CPU?
K.
Keith
It is my overall impression from many POS-reports posted on w:o for germany within the last weeks. (and yes, I am well aware Germany is not the centre of the universe )
K.
wbmw
My impression is that 6xx was designed to replace the 5xx
While I agree that is so for the manufacturing side of the house, it would depend on how many full-cache product they will yield from the new design this year. Actually, my line of thought is Intel needs this design to work their inventory structure back in line over time. Because the capacity to do so is in place anyway, it comes at cost of additional wafers and materials plus yield loss associated with the migration. While the latter is likely the dominant costfactor for H1, it will be hardly visible for accounting reasons: Increasing manufacturing diecost will prevent inventorized die from being devalued, and over time as well from being written off.
Bottomline: It should work out well at the end. But if I am anywhere close with the above background, it would imply 5xx parts around for way longer than most believe. 6xx will be offered for desktop (if only for optical reasons), but it will be premium-priced to ensure proper allocation to make the strategy work for the better part of 05.
K.
Keith
Along the line of drbes reply - I see Intel getting more and more aggressive in valuespace. And - maybe not yet at Gateways, but in many lineups in Europe - Athlon64 playing the dominant role in performance-segment of consumer-lineups.
On a sidenote (on ed's piece) he really seems to believe Intels 5xx line is already done. Which is an utterly misunderstanding imo. I see 6xx overtaking 5xx not anytime soon. I'm not even sure it will happen this year. And 5xx does not have EMT64 capability as well - as pointed out already. Which is all no problem for the corporate fort to defend for Intel. While in consumerspace 5xx will be relegated to value as soon as Win64 is out. Unless Intel will shift gear and make EMT64 5xx as well. Which it probably could from Q3 this year. It will still suck, but nobody will know before 64 apps are out.
K.
Paul
Not so sure. Look what the car-makers do in creating segments: Once we had size and engines as the dominant differentiators. Today it's use-models.
This approach is in the process of transfer into PC-Systems as we speak. Classifying segments with price points only is nothing short of just a historical relict. CPU-model is of a lot less importance than it was couple years ago in this respect anyway. Computing power is already relegated to just one of many features of a CPU in common perception. This all does apply only marginally for the perception of enthusiasts - which we must consider ourselves as well. So we better be aware of a huge skew in our perception of CPUs.
And while we are at introspection and find it hard to understand what Joe Sixpack makes buying this or this model, we should allow the question from time to time if looking for benefits instead of specs is not the better way to look at PCs.
K.
Keith
Yup. Hector's 50x15 vision indicates in what time-horizon he is thinking. I'd venture this applies for CPG as well. It's sort of a marathon man succeeds a sprinter, as bobs pointed it out. Step by step, hill by hill, mile by mile.
And yes, while it's certainly less entertaining than Jerry's wild attack-rides, it's likely more successful.
K.
bobs
Bio diversity is the best insurance a species has to longevity
Sure is. It's just damned hard to explain to the wife
On a more serious note, any comment preventing the board from myopia is highly appreciated. I don't see a third of the market before the very end of this decade for AMD. A fourth by 07 which AMD targets seems doable though. Which, if it is in terms of Dollar share as well, might be sufficient critical mass already to show steady profitability and ensure AMD can afford to pay for leading edge fabbing. Unless their transition to 300mm will be as desastrous as their 130nm bulk shrink, that is.
K.
aleph
I noticed that MSFT are offering a free upgrade to Win64 for those buyers that buy ( future tense ) a 32-bit Winxx.
Afaik this applies (present tense) for the Server version only.
The Client version is free in its preview flavour however - for a year.
And MSFT strongly hinted the Release will be available fairly cheap in bundles for everybody - even bundled with a cable or a mouse or something. They are interested to spread the thing - but dont want to be involved in the driver issues, that's why they do it that way.
K.
Keith
Yeah.
Btw the label inpired me (if it was not the wine)
Opteron [square] for Dual Core?
K.
wbmw
Thanks for all the links. You cannot be blamed for a lack of diligence. But it does not change the world at all. I think it is unnecessary to post the link of Microsofts pullback - let alone digging for other ISV's pullbacks. I am sure you are up-to-date on the topic.
K.
Chipguy
Again, it's all about economies of scale. That is the context you took my quote out of.
Intel itself admitted IPF cannot be brought down to desktop currently - even not for workstations. And Intel admitted IPF cannot compete in low server space as well. So Microsoft's announcement is not surprising at all. I mean if there is no hardware, why should it allocate ressources for an OS?
I am sure you did not miss several IPF-pullbacks of ISV recently as well.
K.
Chris
To begin with, nothing faded away. And no, the announcements expected for Dec 28 are not Turion-related. At least not directly. It is within the same AMD-architecture, but not in its mobile favour.
K.
wmbw
The codebase is already several thousand apps strong, and growing
Several thousand apps is below critical mass. ISV are bailing out. MS does. In fact, codebase is shrinking as we speak.
Economies of scale really kick in if you can sell ten million plus CPUs per node. Everything below is simply not enough to survive in this business.
IPF cannot replace XeonMPs. Codebase.
K.
Keith
Lol. Knock on wood K.
Keith
D'accord. The vast majority of technical issues should be adressed with upcoming platforms and CPU-steppings.
Foodchain-issues likely will prevail. I see these getting tougher this year than we ever have seen it in the past.
K.
chipguy
When I said end of last millenium I meant the timeframe when Merced was able to effectively compute X86 code with its X86 core and therefore allowed a smooth migration. I thought this would be clear from the context. It has certainly nothing to do with Y2K.
And yes, it could be I don't understand technology at all. However, if so it would be irrelevant in the context anyway. I am talking about the broad business picture - and I pointed out already it is not about the performance level achievable with the next generation silicon.
Btw I ask you to care about your wording if you intend to continue the dispute. Thanks.
K.
Joey
While I agree the industry likes competition driving prices down, X86 is not an ideal architecture for the top-end space.
IPF is far more ready to occupy this space. But, as you are mentioning dynamics, what I say is a static argument which may be invalid in some five years when HORUS 3 or the like might be able to bridge the gap.
K.
Keith
Yes. It's because Ati IPP is still very limited in volume. Will take some months until its unrestrictedly available. And I should add again I remain sceptical if all what ATI announced will work instantly - from not always satisfying experience with this company's products so far.
Anyway, a lot of things should fall into places by the time of Win-64 release, I hope.
I would not write AMDs notebooks off for this year already. It could well be premature. 64-bit notebooks could easily become the blockbusters of this year's back to school season and going forward. (The driver considerations for periphery might play a less dominant role in this segment). I would not at all be surprised if the mobiles share would be equivalent to the desktop-share going out of this year.
K.
Chipguy
I guess I have pointed out several times already that i consider IPF as a superior architecture over X86. However it missed its window of opportunity to move the world - which was clearly open until the end of the last millenium. It is closed now and cannot be reopened. This is not at all because of hardware. Its because of codebase. So whether Montecito will be a masterpiece of silicon or just another chip - it does not make any difference.
The line of arguments is about as follows:
No critical mass of codebase --> no critical mass of manufacturung volume ---> Not enough economies of scale in manufacturing ---> no oppoerutity to reach desktops ---> not enough Cash Flow to allocate to development ---> not enough support from ecosystem partners ---> No hope to make any money at any time with it for Intel.
Bottomline: There is nothing that would prevent IPF from goint the Alpha-way (which is far superior over X86 as well from an architectural vantagepoint as well).
---> It needs an exit strategy. Which you better let look as another migration, so you gotta make sure you have something to show by this time. Likely an eight-core fast X86-64 Chip outperforming the current four-core Itanium thing, as the world would not buy your next big thing anyway. And you need to build bridges to save face. And plan the timing properly - or (pardon) at least better as the entry phase.
That is what Intel is doing as we post. For good. It cannot be done today, but in some three years all the pieces they need should be ready.
K.
i_banker
I look at the common platform as part of an exit strategy of IPF: It's the golden hardware brigde to migrate to X86. It would need a software bridge as well, which is easy to imagine already: We have seen a software emulation to run X86 on IPF. Same is possible to run EPIC on X86. It's basically all nonsense nobody with a sane mind would ever use, but it will prevent Intel from loosing face when exiting IPF.
K.
Keith
AMD never claimed so. (Or maybe I better should say New AMD never claimed so )
The 50M claim came from Mike Masdea - a misunderstanding Bob cleared up when MM confronted him with it publicly at CSFB conference. Bob was diplomatic enough to say under (theoetical) circumstances AMD would be able to manufacture 50m CPUs in O5 - and added the actual number could be lower as AMD migrates to Dual-Core etc.
K.
Keith
There is an issue on Ati's Southbridge. If you combine Ati's Northbridge with ULi Southbridge until Ati fixed the issue everything is fine.
K.
Combjelly
I don't know either - anyway, these duties were intended to spur local assembly creating local jobs, so they likely persist.
K.
CJ
Thanks. Now, twenty percent sounds like a good argument for a local assembly plant.
K.
i_banker
Well. See my post to buggi.
K.
Buggi
Well. Not for charity. If i'd be interested in what they have to offer, I'd prefer a favourable give and take transaction to enable them to continue their development over the circling vulture strategy.
K.